Reconstruction of high-quality MR images requires precise knowledge of the dynamic gradient magnetic fields used to perform spatial encoding. System delays and eddy currents can perturb the gradient fields in both time and space and significantly degrade the image quality for acquisitions with an ultrashort echo time or with rapidly varying readout gradient waveforms. A technique for simultaneously characterizing and correcting the system delay and linear-and zero-order eddy currents of an MR system is proposed. A single set of calibration scans were used to compute a set of system constants that describe the effects of system delays and eddy currents to enable accurate reconstruction of data collected before uncorrected eddy currents have decayed. Accurate MR image formation requires precise control of the gradient magnetic fields to traverse a defined k-space trajectory during data acquisition. System delays and eddy currents can cause the actual trajectory to deviate substantially from the ideal trajectory (1,2). Modern MRI systems include preemphasis hardware for eddy current compensation (3-5) that correct linear (first-order) eddy currents by altering the digital gradient waveform (1) or high-pass filtering the gradient waveform (6) and correct B 0 (zero-order) eddy currents by modulating the receiver phase (5) or using a current-controlled B 0 coil. Because preemphasis systems are limited to a small number of time-constants, it is not uncommon for short-time constant eddy currents to remain even when they are optimally configured. System delays (e.g., amplifier group delay) are typically determined during manufacturing or installation using calibration scans and manual tuning. However, because eddy currents with time-constants shorter than, or comparable to, the echo-time cannot be distinguished from a true delay (1), the measured delay value is more properly termed the apparent system delay.Traditional Cartesian acquisitions are inherently robust to uncorrected delays and short time-constant eddy currents, particularly when the echo-time is long and the data are primarily acquired under a constant gradient amplitude after all uncorrected eddy currents have decayed. These conditions are often met when preemphasis is used to compensate long time-constant eddy currents, enabling high-quality images to be obtained with minor retrospective data corrections. In contrast, acquisitions that use an ultrashort TE (UTE) or traverse complex k-space trajectories using high slew-rate, time-varying readout gradient waveforms (e.g., Twisted Projection Imaging [TPI], 3D Cones) sample data while uncorrected eddy currents are present, distorting the actual k-space path, and may have an undefined center of k-space due to the unknown system delay, leading to significant image degradation (2).A potential solution is to measure the actual k-space trajectories (7-12) and eddy currents (12) and to use them to accurately reconstruct the acquired data. While such measurement approaches work well in principle, all unique gradient w...